c 2018 by robert g. littlejohn - apache2 ubuntu...
TRANSCRIPT
Copyright c© 2018 by Robert G. Littlejohn
Physics 221B
Spring 2018
Notes 35
Introduction to Scattering Theory
and Scattering from Central Force Potentials
1. Introduction
Scattering is a fundamental physical process that is important in a wide range of applications.
For example, most experimental work in particle physics involves scattering, and a large part of
what is known about the properties of particles, both elementary and composite, has been obtained
through scattering experiments. For another example, scattering of nuclear species in a hot plasma
is important in determining thermonuclear reaction rates in stars or in the early universe. For
yet another, one of the current experimental techniques for investigating carbon nanotubes is the
scattering of photons. A long list of examples could be accumulated.
We have previously considered some scattering processes from the standpoint of time-dependent
perturbation theory, in Notes 33 and 34. There the general idea was to take a simple initial condi-
tion describing the incident particle (for example, a plane wave), and to solve the time-dependent
Schrodinger equation to find transitions to final states in which there is a scattered particle. In these
notes we begin a more systematic study of scattering theory, taking a mostly time-independent ap-
proach. That is, we will solve the time-independent Schrodinger equation, looking for scattering
solutions of a given energy. As we will see, there is an interplay in scattering and decay processes
between the time-dependent and the time-independent approaches.
2. Classical and Quantum Potential Scattering
In this introduction we will concentrate mostly on potential scattering in three dimensions, that
is, the scattering of particles from a localized potential V = V (x). The potential should go to zero
as r = |x| → ∞; this is normally understood in scattering theory. Initially we will take a somewhat
intuitive approach to the theoretical development, but some of the mathematical arguments are
simpler if we assume that the potential rigorously vanishes outside some cut-off radius r = rco. If
the potential is cut off in this way then the wave function ψ(x) satisfies the free particle Schrodinger
equation for r > rco.
We will draw certain conclusions about the asymptotic form of the wave function when the
potential rigorously vanishes for r > rco, but then the question arises as to whether these conclusions
still hold for realistic potentials that only go to zero gradually as r → ∞. We will address this
question in Sec. 8, where we will find that the main conclusions still hold as long as the potential
2 Notes 35: Introduction to Scattering Theory
goes to zero faster than 1/r as r → ∞. Of course, the potential that goes to zero precisely as 1/r is
the Coulomb potential, which is important in applications. Its treatment, however, requires special
techniques which (some day) will appear in a separate set of notes. In the meantime, a discussion
of Coulomb scattering may be found in Schiff’s or Sakurai’s book on quantum mechanics.
We will speak of the scattering of a particle by a potential V (x), as if it were a fixed potential in
space, but in reality the beam particle interacts with a target, which often is another particle. Then
a proper treatment requires us to take into account the dynamics of both particles. The changes
necessary to do this were discussed in some detail in Notes 33. For now we simply note that if the
target is very massive, then it can be thought of as producing a potential for the beam particle that
is fixed in space (that is, in an inertial frame), as we shall assume in these notes.
For simplicity, we will assume that there are no spin-dependent forces, so that the Hamiltonian
for the scattering problem has the kinetic-plus-potential form.
p
x
y
z
V (x)
Fig. 1. Classical scattering. A beam of particles, all with the same momentum p, is directed against a target.
To gain insight into scattering in quantum mechanics, we first consider classical scattering,
which is illustrated in Fig. 1. A beam of particles, all of the same momentum p, is directed against
a target. When the beam is first turned on, it takes some time for the beam to reach the target,
after which particles are scattered in various directions. Potential scattering is elastic, that is, when
particles gain kinetic energy by falling into a potential well, they lose the same amount when climbing
back out again, so the kinetic energy of the scattered particles at a large distance from the scatterer
is the same as the kinetic energy in the incident beam. Only the direction of motion has changed.
If we sit at an observation point some distance from the scatterer, it takes some time for the first
scattered particles to reach us, after which the flux of particles coming from the scatterer reaches a
steady state.
Let us now consider how this picture changes in quantum scattering. In quantum mechanics a
state of a definite momentum p is a plane wave Ceik·x, where C is a constant and p = hk, so it is
natural to associate the incident beam in a scattering experiment with a plane wave. This involves
some idealization, since a real plane wave fills up all of space and, in particular, it extends to infinity
in the transverse direction, while real beams are always of finite extent in the transverse direction.
Notes 35: Introduction to Scattering Theory 3
As in our picture of classical scattering, let us imagine that the beam is turned on at some
time, that is, the plane wave is cut off in the longitudinal direction, with a front that advances from
the source (at a large distance from the scatterer) toward the scatterer. When the wave reaches
the scatterer it interacts with it, producing a scattered wave which radiates out in all directions, as
illustrated in Fig. 2. If we sit at an observation point at some distance from the scatterer, then,
as in the case of classical scattering, after a while a steady state is reached, in which there is a
definite flux of scattered particles at our observation point. Observers further out will have to
wait longer for the steady state to be reached, but after sufficient time a steady state of scattered
particles is reached at any fixed observation point. Meanwhile parts of the incident wave that have
not interacted with the scatterer (speaking somewhat loosely) continue forward, progressively filling
space in the longitudinal direction with the incident wave. Speaking a little more precisely, since
we are imagining that the incident wave fills all of space in the transverse direction, we can imagine
that as it passes around the scatterer, the scatterer effectively punches a hole in the incident beam,
that is, it creates a shadow, into which the incident beam will gradually diffract as it proceeds
downstream. The diffracted waves have their own flux which is not exactly in the same direction
as the incident wave, and so must be regarded as representing scattered particles. In any case, the
fact remains that at any given distance from the scatterer, including in the forward direction, there
is some time after which a steady state is reached.
p = hk
x
y
z
V (x)
Scattered Wave
Incident Wave
Fig. 2. Quantum scattering by a potential V (x). Incident and scattered waves are shown.
In quantum mechanics an energy eigenstate, a solution of the time-independent Schrodinger
equation of definite energy E, is sometimes called a “stationary state,” because the probability
density ρ = |ψ|2 and the probability current J (see Secs. 5.14 and 5.15) are independent of time. In
the picture of scattering described above, the initial state, a truncated plane wave approaching the
scatterer from some distance, is not a stationary state, because it has a nontrivial time evolution.
Nevertheless, the intuitive picture suggests that if we sit at some observation point x and wait long
enough, then the wave function Ψ(x, t) will approach an energy eigenfunction, that is, it will take
on the form ψE(x)e−iEt/h in the neighborhood of our observation point, where ψE(x) is a solution
of the time-independent Schrodinger equation of energy E.
Recall that in the analysis of scattering in Notes 33 we took an initial state (a plane wave filling
4 Notes 35: Introduction to Scattering Theory
all of space) that was not an energy eigenstate of the total Hamiltonian H0 +H1 = p2/2m+ V (x),
only an eigenstate of the unperturbed HamiltonianH0 = p2/2m. Then we studied the time evolution
of this initial state under the full Hamiltonian, and derived from it the scattering rate and from that
the cross section. Although the initial state considered in Notes 33 is not the same as the truncated
plane wave we are considering here, nevertheless in both cases the same idea applies, that an energy
eigenstate can be extracted from the long time evolution of the initial state.
3. The Time-Independent Approach
A more direct approach, however, which we shall pursue in the rest of these notes, is simply to
look for an energy eigenstate, that is, a solution of the time-independent Schrodinger equation,
−h2
2m∇2ψ(x) + V (x)ψ(x) = Eψ(x). (1)
The solution should have the right boundary conditions at large distances from the scatterer to
represent both the incident plane wave and the scattered wave. At distances outside the cut-off
of the potential, that is, when r > rco, the potential vanishes (as we are assuming) and the wave
function ψ(x) must satisfy the free particle Schrodinger equation with energy E. In particular,
this must apply to both the incident wave and the scattered wave. Since the incident wave is a
plane wave, which is already a solution of the free particle Schrodinger equation, we must have
E = h2k2/2m, that is, the energy eigenvalue in the Schrodinger equation (1) is the same as the
kinetic energy of the incident beam. Since the experimenter is free to give the incident particles
any positive kinetic energy, we see that E is not quantized, that is, the scattering solutions that we
shall find belong to the continuous spectrum with E > 0. Note, however, that the energy E that
appears in (1) applies at all points of space, even where the potential is nonzero. Wave functions of
the continuous spectrum are not normalizable, which is reasonable in the present case because the
incident plane wave is not normalizable.
The intuitive picture we have developed suggests that there is a unique solution of the Schro-
dinger equation (1) of energy E > 0, given the momentum of the incident plane wave p. This must
be the energy eigenfunction that is obtained in the long time limit from the time-dependent solution
when a truncated plane wave is directed at the target, as discussed above. This intuitive idea can
be translated into a theorem, that there is a unique solution of the time-independent Schrodinger
equation (1) of energy E > 0, given that the boundary conditions at r → ∞ include the incident
plane wave of given momentum p and outgoing scattered waves. We will see more rigorously how
this unique solution is obtained in the case of central force potentials in this set of notes, and in the
case of arbitrary potentials (and other generalizations) in subsequent notes.
However, the solution so obtained is degenerate, that is, for the given E there is more than one
linearly independent solution of the Schrodinger equation (1). This follows simply from the fact that
we can change the direction of p (the direction of the incident beam) without changing E, thereby
generating a class of linearly independent solutions of the same energy. In three dimensions, these
Notes 35: Introduction to Scattering Theory 5
solutions are parameterized by the direction of the incident momentum, so there is a continuous
infinity of them. Note that in one dimension, there are only two linearly independent solutions,
corresponding to an incident wave directed at the scatterer from the right or from the left.
4. Boundary Conditions
To be more precise about the incident and scattered waves, let ψ(x) be an exact solution of the
Schrodinger equation (1) of energy E, and define the incident wave by
ψinc(x) = eik·x, (2)
where the wave vector k is related to the energy eigenvalue in the Schrodinger equation by the free
particle relation,
E =h2k2
2m. (3)
The incident wave does not satisfy the Schrodinger equation (1) everywhere, only in the region
r > rco where the potential vanishes. Then define the scattered wave by
ψscatt(x) = ψ(x)− ψinc(x), (4)
so that
ψ(x) = ψinc(x) + ψscatt(x), (5)
exactly. Notice that since ψ satisfies the Schrodinger equation exactly, and since ψinc does also in
the region r > rco where V = 0, then so does ψscatt in this region. But neither ψinc nor ψscatt
satisfies the Schrodinger equation (1) in the scattering region, where V 6= 0.
We will be interested in the asymptotic form of the scattered wave, that is, when r → ∞,
because this is where the experimenter will locate the detection apparatus. In particular, r → ∞
means r ≫ rco (continuing with the simplification that V = 0 for r > rco).
It is possible to guess the asymptotic form of the scattered wave. The intuitive picture we have
developed suggests that in the asymptotic region the scattered wave will have the form of a spherical
wave traveling outward from the scatterer. At distances r ≫ rco the scatterer subtends only a small
solid angle as seen from an observation point, and so it defines an almost unique direction in which
the waves must be propagating (that is, away from the scatterer, in the radial direction). Since
the probability density in quantum mechanics is |ψ|2, we expect the scattered wave to fall off with
distance r from the scatterer as 1/r2, since as the scattered particles travel outward they are spread
over a sphere of increasing radius r and area proportional to r2. This means that the scattered
wave function ψscatt should fall off as 1/r. This is only an asymptotic form, that is, it is the leading
behavior of ψscatt in an asymptotic expansion. In addition, we must allow ψscatt to have an angular
dependence to account for the different scattering rates in different directions. Finally, we require it
to be proportional to eikr, since this (when multiplied by the time dependence e−iEt/h) produces an
6 Notes 35: Introduction to Scattering Theory
outwardly propagating spherical wave. Overall, we guess that the asymptotic form of the scattered
wave is
ψscatt(x) ∼eikr
rf(θ, φ), (6)
where the function f(θ, φ) accounts for the angular dependence.
We emphasize that this is only an asymptotic form, valid when r is large. In these notes the
symbol ∼, when used in an equation such as (6), will mean that the left hand side equals the right
hand side, plus terms that go to zero faster than the right hand side as r → ∞. In other words, the
right hand side is the leading term of an asymptotic expansion of the left hand side. Notice that
the k parameter in the asymptotic form of the scattered wave is the same as |k|, where k is the
wave number appearing in the incident wave (2). This corresponds to the fact that the scattering is
elastic, that is, the classical fact that in the asymptotic region the scattered particles have the same
kinetic energy as the incident particles.
We should check that our guess (6) for the asymptotic form of the scattered wave satisfies the
Schrodinger equation at large r, at least to leading order in powers of 1/r. Since the potential is
negligible at large r we are talking about the free-particle Schrodinger equation. We see that it does
if we compute the Laplacian of (6),
∇2[eikr
rf(θ, φ)
]
= −k2[eikr
rf(θ, φ)
]
+O( 1
r3
)
, (7)
where the dominant term comes from the radial term of the Laplacian in spherical coordinates (see
Eq. (D.23)). Thus the kinetic energy term in the Schrodinger equation balances the total energy
term Eψ, to leading order in 1/r.
5. The Scattering Amplitude and Cross Section
The function f(θ, φ) is called the scattering amplitude. It is in general a complex function of
the angles. A good deal of scattering theory is devoted to determining the scattering amplitude,
since it bears a simple relation to the differential cross section.
y
z
V (x)
∆Ω
Detector
x
Fig. 3. A particle detector intercepts a small solid angle ∆Ω which defines a cone as seen from the scatterer. Particlesscattered into this cone will enter the detector.
Notes 35: Introduction to Scattering Theory 7
Cross sections and differential cross sections were defined in Notes 33, where the relation between
cross sections and transition rates was discussed. In the present case let us imagine an experimental
situation such as illustrated in Fig. 3. A detector, located at some distance from the scatterer,
intercepts all scattered particles coming out in a small cone of solid angle ∆Ω, centered on some
direction n = (θ, φ). The counting rate is the scattered current Jscatt, integrated across the aperture
to the detector.
The probability density ρ = |ψ|2 and probability current J were discussed in Secs. 5.14 and 5.15.
Note that the expression for J depends on the form of the Hamiltonian. In the present context,
we are interested in Hamiltonians of the kinetic-plus-potential type, for which the current can be
written
J =h
mIm(ψ∗∇ψ), (8)
which is a version of Eqs. (5.56) and (5.57).
In Notes 5, we were thinking of normalized wave functions, for which ρ and J are interpreted as
the probability density and current. In the present context, however, the wave functions are scat-
tering solutions that are not normalizable. It is best in this context to think of ρ and J as a particle
density and current (with dimensions of particles/volume and particles/area-time, respectively).
The particle density in the incident beam is |ψinc|2, or,
ninc = 1. (9)
There is one particle per unit volume in the incident beam. Similarly, the incident current is easy
to compute. It is
Jinc =hk
m= v, (10)
where v is the velocity of the incident beam. In magnitude, Jinc = v.
We will also need the scattered current to get the counting rate. We first compute the gradient
of the scattered wave (6) in spherical coordinates (see Eq. (D.20)),
∇ψscatt(x) = r[ ik eikr
rf(θ, φ)
]
+O( 1
r2
)
, (11)
where we only carry the result to leading order in 1/r. Substituting this into (8), we find
Jscatt ∼hk
m
|f(θ, φ)|2
r2r. (12)
To leading order, the asymptotic form of the scattered current is purely in the outward radial
direction, as we expect. Terms in Jscatt that go to zero faster than 1/r2 as r → ∞ will not contribute
to the counting rate, since the area of the aperture to the detector, for fixed ∆Ω, is proportional to
r2.
Now we integrate the scattered current over the aperture to the detector, which we assume is
at large distance r from the scatterer, and which has area r2∆Ω. Then the counting rate is
dw
dΩ∆Ω =
∫
aperture
Jscatt · da = v|f(θ, φ)|2 ∆Ω, (13)
8 Notes 35: Introduction to Scattering Theory
where da is the area element of the aperture and where we have used v = hk/m. Now dividing this
by Jinc = v, we obtain
dσ
dΩ= |f(θ, φ)|2, (14)
a simple result. This equation explains the interest in finding f(θ, φ), since it leads immediately to
the differential cross section, which is measurable. Notice that the counting rate, which is physically
observable in a scattering experiment, depends only on the leading asymptotic form of the scattered
wave function. One does not need to know the exact form of the scattered wave (at smaller values
of r where correction terms to the asymptotic expansion of the scattered wave are important, or in
the scattering region where the potential is nonzero).
Now for some comments about this calculation. Notice that we computed the flux of the
scattered particles intercepted by the detector, but we ignored the incident particles. In realistic
scattering experiments the detector is usually located outside the beam, so incident particles do not
enter it. Our incident wave (2) fills all of space, but that is an artifact of our simplified model. In
reality the beam will be cut off at some finite transverse size, and if the scattering angle is not too
small, the detector can be located outside the beam.
If the scattering angle is very small, however, then the detector will have to be inside the beam
and it will detect incident particles as well as scattered ones. The counting rate is still the particle
flux intercepted by the aperture of the detector, but the flux is neither the incident flux nor the
scattered flux nor the sum of the two, but rather the flux computed from the total wave function
ψ = ψinc + ψscatt. Since the current is quadratic in the wave function, there are cross terms or
interference terms between the incident wave and the scattered wave. Taking all these effects into
account leads to the optical theorem, which we take up in Sec. 14.
6. Central Force Scattering
We now specialize to the important case of a central force potential, V = V (r). The main
simplification in this case is that the Schrodinger equation (1) is separable in spherical coordinates
(see Notes 16), so the solutions can be written in the form
ψkℓm(x) = ψkℓm(r, θ, φ) = Rkℓ(r)Yℓm(θ, φ), (15)
where Rkℓ(r) is a solution of version one of the radial Schrodinger equation, Eq. (16.7). The radial
eigenfunction Rkℓ(r) is parameterized by ℓ because the radial Schrodinger equation contains ℓ in the
centrifugal potential. It is also parameterized by the energy E, which we will normally indicate by an
equivalent wave number k, given by E = h2k2/2m. The three-dimensional solution ψ depends on k,
ℓ andm, as indicated in Eq. (15). The radial eigenfunction Rkℓ(r) is related to an alternative version
of the radial eigenfunction ukℓ(r) = rRkℓ(r), which satisfies version two of the radial Schrodinger
equation, Eq. (16.11). (Function u was called f in Notes 16.)
Notes 35: Introduction to Scattering Theory 9
For our present purposes we rearrange these two versions of the radial Schrodinger equation as
1
r2d
dr
(
r2dRkℓ
dr
)
+ k2Rkℓ(r) =W (r)Rkℓ(r) (16)
and
u′′kℓ(r) + k2ukℓ(r) =W (r)ukℓ(r), (17)
where
W (r) =ℓ(ℓ+ 1)
r2+
2m
h2V (r). (18)
The solution ψkℓm of the Schrodinger equation is a simultaneous eigenfunction of (H,L2, Lz),
with quantum numbers (kℓm), of which k > 0 is continuous and ℓ and m are discrete. This solution
does not satisfy the boundary conditions we require for a scattering problem. Such solutions do,
however, form a complete set, so any solution of a given energy E can be expressed as a linear
combinations of solutions of the form (15) of the same value of E. That is, the most general solution
of the Schrodinger equation of a given E = h2k2/2m > 0 is given by
ψ(x) =∑
ℓm
CℓmRkℓ(r)Yℓm(θ, φ), (19)
where Cℓm are the expansion coefficients. The freedom in the choice of the coefficients Cℓm indicates
that there is a large degeneracy in the solutions of the Schrodinger equation of a given energy
E > 0, a fact that we have already noted. In moment we will find the expansion coefficients for our
desired scattering solution, in terms of some simple parameters that characterize the potential (the
parameters turn out to be the phase shifts δℓ).
7. Free Particle Solutions
In the special case of the free particle, V (r) = 0, the radial wave function Rkℓ(r) is a linear com-
bination of the two types of spherical Bessel functions, jℓ(kr) and yℓ(kr), as discussed in Secs. 16.6
–16.8. You may wish to review the contents of those sections of Notes 16. Here we just note a
slightly modified version of Eqs. (16.29) and (16.30),
jℓ(ρ) ≈1
ρsin
(
ρ− ℓπ
2
)
, (20)
yℓ(ρ) ≈ −1
ρcos
(
ρ− ℓπ
2
)
, (21)
valid when ρ≫ ℓ.
Sometimes we are interested in solutions for a particle that is free everywhere, that is, where
the potential V (r) = 0 for all r, all the way down to r = 0. In that case the y-type Bessel functions
are not allowed, and Rkℓ(r) = jℓ(kr). Then the most general free particle solution of energy E is a
sum of the form (19), that is,
ψfree(x) =∑
ℓm
Cℓm jℓ(kr)Yℓm(θ, φ), (22)
10 Notes 35: Introduction to Scattering Theory
for some expansion coefficients Cℓm.
In particular, it must be possible to express the incident wave eik·x in this form. The problem of
determining the expansion coefficients Cℓm in this case is discussed in Sakurai’s book, and will not be
repeated here. The derivation involves some use of recursion relations satisfied by the spherical Bessel
functions and the Legendre polynomials, and is straightforward but not particularly illuminating. (A
more illuminating derivation uses group theory, which, however, is outside the scope of this course.)
The result, however, is very useful. It is
eik·x = 4π∑
ℓm
iℓjℓ(kr)Y∗ℓm(k)Yℓm(r), (23)
where r and k refer to the directions of the vectors x and k on the left hand side, and where Yℓm(r)
means the same as Yℓm(θ, φ). This expansion can be rewritten with the use of the addition theorem
for spherical harmonics, Eq. (15.71), which for present purposes we write in the form
Pℓ(r · k) = Pℓ(cos γ) =4π
2ℓ+ 1
∑
m
Y ∗ℓm(k)Yℓm(r), (24)
where γ is the angle between x and k. Then the expansion (23) becomes
eik·x =∑
ℓ
iℓ(2ℓ+ 1)jℓ(kr)Pℓ(cos γ). (25)
Notice that if we take k to lie in the z-direction, k = kz, then γ is the same as θ, the usual spherical
coordinate. In that case, the incident plane wave becomes
eik·x = eikr cos θ, (26)
that is, it is independent of the azimuthal angle φ, and the sum (25) is the expansion of the plane
wave in Legendre polynomials of cos θ.
8. Asymptotic Form of the Radial Wave Functions
If the potential V (r) is nonzero but vanishes outside a cut-off radius r = rco, then the radial
eigenfunctions Rkℓ(r) in the region r > rco are linear combinations of both jℓ(kr) and yℓ(kr). The yℓ-
type solutions are allowed because there is no attempt to extend them down to r = 0. Then, in view
of Eqs. (18) and (19), we can write the leading asymptotic form of Rkℓ(r) as a linear combination
of two functions,
Rkℓ(r) ∼sin(kr − ℓπ/2)
kr,
cos(kr − ℓπ/2)
kr. (27)
Equivalently, the leading asymptotic form of ukℓ(r) is a linear combination of two simple exponen-
tials,
ukℓ(r) ∼ e±ikr . (28)
What about when the potential does not cut off at a finite radius, but goes to zero gradually?
One might guess that if it goes to zero fast enough, then when r → ∞ the potential would not
Notes 35: Introduction to Scattering Theory 11
matter, and the leading asymptotic forms (27) and (28) would still be valid. Let us examine the
asymptotic form of the radial wave functions more carefully to see if this is true.
We now allow V (r) to go to zero gradually as r → ∞, with no rigorous cut-off. We cannot find
an explicit form of the radial eigenfunctions Rkℓ(r) or ukℓ(r) without knowing the potential and
solving the radial Schrodinger equation, but determining just the asymptotic form does not require
this. It is easier to work with the u-form of the radial Schrodinger equation (17) for this analysis.
Since both the true potential and the centrifugal potential go to zero as r → ∞, the function
W (r), defined by Eq. (18), also goes to zero as r → ∞. So as a first stab at analyzing the asymptotic
form of ukℓ(r), let us neglect W (r) altogether on the right hand side of the radial Schrodinger
equation (17). Then the solution for ukℓ(r) is simply a linear combination of the exponentials (28).
To take into account the effects of W (r), let us write the solution of Eq. (17) as
ukℓ(r) = eg(r)±ikr, (29)
where g(r) is a correction that will account for the effects of the function W (r). If we can show that
g(r) → 0 as r → ∞, then the correct asymptotic forms of ukℓ(r) will be a linear combination of
the solutions (28). That is, the function W (r) on the right hand side will make no difference in the
asymptotic form of the solution. Substituting Eq. (29) into the radial Schrodinger equation (17), we
obtain an equation for g,
g′′ + g′ 2 ± 2ikg′ =W (r). (30)
This is rigorously equivalent to Eq. (17).
Now we consider the behavior of W (r) as r → ∞. If the true potential V (r) goes to zero faster
than 1/r2, then W (r) is dominated by the centrifugal potential and also goes to zero as 1/r2. We
assume ℓ 6= 0, so the centrifugal potential does not vanish. If the true potential goes to zero more
slowly than 1/r2, let us assume that it does so as a power law, 1/rp, where 1 < p ≤ 2. This excludes
the case of the Coulomb potential, for which p = 1, but it approaches the Coulomb potential as
p→ 1. Then W (r) has the asymptotic form,
W (r) ∼a
rp, (31)
where 1 < p ≤ 2 and a is a constant. The case ℓ = 0 can be handled as a special case, but it does
not change any of the conclusions that we shall draw.
We are interested in solving Eq. (30) in the asymptotic region when W has the asymptotic
behavior (31). Let us assume that g is asymptotically given by a power law,
g(r) ∼b
rs, (32)
where b is another constant and s > 0. Then the three terms on the left hand side of Eq. (30) go as
1/rs+2, 1/r2s+2 and 1/rs+1, in that order. The third term dominates, so to leading order we have
s = p− 1, so that 0 < s ≤ 1, and the power law ansatz for g has a solution. That is, g(r) ∼ b/rp−1,
12 Notes 35: Introduction to Scattering Theory
g(r) does go to zero as r → ∞, and thus the asymptotic form of ukℓ(r) is a linear combination of
e±ikr.
But in the case of the Coulomb potential, W (r) ∼ a/r, that is, with p = 1. Then taking the
dominant term in Eq. (30), we have
±2ikg′(r) =a
r, (33)
or,
g(r) = ∓ia
2kln(kr). (34)
The logarithm does not go to zero as r → ∞, in fact, it increases without bound (although very
slowly). Thus, in the case of the Coulomb potential, the asymptotic form of the radial wave function
is
ukℓ(r) ∼ exp
±i[
kr −a
2kln(kr)
]
. (35)
The Coulomb potential gives rise to long range, logarithmic phase shifts that do not approach the
free particle phases as r → ∞.
In summary, if the potential V (r) goes to zero faster than 1/r, then the asymptotic form of the
radial wave function is given by the linear combinations shown in Eq. (27) (for Rkℓ) or in Eq. (28)
(for ukℓ). For the rest of these notes we will assume that V (r) does satisfy this condition, thereby
excluding the Coulomb potential.
9. Partial Waves
With this assumption about V (r), we can write the asymptotic form of Rkl(r) as a linear
combination of the functions (27), that is,
Rkℓ(r) ∼A sin(kr − ℓπ/2) + B cos(kr − ℓπ/2)
kr, (36)
where A and B are constants. Since the radial Schrodinger equation (16) is a real equation, we can
choose the functions Rkℓ(r) to be real, and thus the coefficients A and B are real. We have not spec-
ified the normalization we will use for the radial wave function, but if we change the normalization
it amounts to multiplying A and B by a common constant. The ratio A/B or B/A, however, can
only be determined by solving the radial Schrodinger equation.
We multiply Eq. (36) by a constant to make the new values of A and B satisfy (A2+B2)1/2 = 1.
Then by using trigonometric identities Eq. (36) can be written
Rkℓ(r) ∼sin(kr − ℓπ/2 + δℓ)
kr, (37)
where cos δℓ = A, sin δℓ = B. The phase δℓ depends on both ℓ and k, but we suppress the k- (or
energy-) dependence in the notation.
The resulting asymptotic form has a single parameter, the phase shift δℓ. Knowledge of δℓ is
equivalent to knowledge of the ratio A/B or B/A. We should compare the asymptotic form (37),
Notes 35: Introduction to Scattering Theory 13
which applies in the case of a potential V (r), to the asymptotic form for the free particle, for which
Rkℓ(r) = jℓ(kr). Using Eq. (20), we can write the latter as
Rkℓ(r) ∼sin(kr − ℓπ/2)
kr(free particle). (38)
Thus the phase shift δℓ in Eq. (37) is measured relative to the free particle phase.
The phases that occur in the asymptotic forms of the radial wave function in the case of any V (r),
Eq. (37), and in the case of the free particle, Eq. (38), can be visualized as follows. The solutions in
both cases apply in the asymptotic region (large r), and both solutions are linear combinations of
the spherical waves e±ikr/r, of which e−ikr/r is inward traveling, and eikr/r is outward traveling.
We can imagine sending in an inward traveling spherical wave from a large distance, which collapses
toward r = 0, then bounces back out as an outward going spherical wave. We can then measure the
phase difference between the inward and outward going waves. We get a certain phase difference in
the case of a free particle, and a different one in the case of the potential. The effect of the potential
is to introduce a phase shift in the reflected wave, which is measured by δℓ.
Now let us return to Eq. (19), the expansion of a general solution of the Schrodinger equation
of energy E as a linear combination of eigenfunctions of H , L2 and Lz, and let us split a factor of
4πiℓ from the coefficients Cℓm in order to make the sum look more like the plane wave expansion,
Eq. (23). That is, let us write the general solution as
ψ(x) = 4π∑
ℓm
iℓCℓmRkℓ(r)Yℓm(r). (39)
We wish to determine the coefficients Cℓm such that the solution (39) will incorporate the incident
plane wave and outgoing scattered wave of the desired scattering solution.
Interpreting (39) as the desired scattering solution of the Schrodinger equation (1), we subtract
Eq. (23) from this and use Eq. (4), obtaining an expansion of the scattered wave,
ψscatt(x) = ψ(x) − eik·r = 4π∑
ℓm
iℓ[
CℓmRkℓ(r)− jℓ(kr)Y∗ℓm(k)
]
Yℓm(r). (40)
At large distances, the scattered wave must consist of purely outgoing spherical waves. Taking the
large r limit and using Eqs. (37) and (38), the quantity in the square brackets in Eq. (40) becomes
[. . .] =1
kr
[
Cℓm sin(kr − ℓπ/2 + δℓ)− Y ∗ℓm(k) sin(kr − ℓπ/2)
]
. (41)
This is a linear combination of incoming and outgoing waves that are proportional to e−ikr/r and
e+ikr/r, respectively. The incoming part is
−1
kr
1
2i
[
Cℓm e−i(kr−ℓπ/2+δℓ) − Y ∗ℓm(k) e−i(kr−ℓπ/2)
]
. (42)
But the scattered wave must be purely outgoing, so this must vanish. This implies
Cℓm = eiδℓ Y ∗ℓm(k). (43)
14 Notes 35: Introduction to Scattering Theory
We see that the exact expansion coefficients of the scattering solution (39) can be determined
in terms of the asymptotic phase shifts of the radial eigenfunctions. Now substituting this back into
the outgoing part of Eq. (41), we find, for the quantity in the square brackets in Eq. (40),
[. . .] =1
krY ∗ℓm(k) ei(kr−ℓπ/2)
(e2iδℓ − 1
2i
)
. (44)
Then the asymptotic form of the scattered wave (40) becomes
ψscatt(x) ∼ 4πeikr
kr
∑
ℓm
eiδℓ sin δℓY∗ℓm(k)Yℓm(r), (45)
where we have used iℓ e−iℓπ/2 = 1. Using the addition theorem for spherical harmonics (24), this
can also be written,
ψscatt(x) ∼eikr
kr
∑
ℓ
(2ℓ+ 1)eiδℓ sin δℓ Pℓ(cos θ), (46)
where we have assumed that k = kz so that the angle between k and r is the usual spherical angle θ.
Since k is the direction of the incident beam, and since r is the direction to some distant observation
point, θ can be interpreted as the scattering angle.
Equation (46) is called the partial wave expansion of the scattered wave, that is, its angular
dependence is expanded in spherical harmonics or Legendre polynomials with coefficients given in
terms of the asymptotic phase shifts of the radial eigenfunctions. Comparing this with Eq. (6) we
obtain the partial wave expansion of the scattering amplitude,
f(θ, φ) =4π
k
∑
ℓm
eiδℓ sin δℓ Y∗ℓm(k)Yℓm(θ, φ), (47)
or, setting k = kz and using the addition theorem for spherical harmonics,
f(θ) =1
k
∑
ℓ
(2ℓ+ 1)eiδℓ sin δℓ Pℓ(cos θ). (48)
In the latter case it is obvious from symmetry that the scattered wave and hence the scattering
amplitude are independent of the azimuthal angle φ, as indicated by the expansion (48). This is a
consequence of the rotational invariance of the potential V (r).
Squaring the scattering amplitude we obtain the differential cross section,
dσ
dΩ(θ, φ) =
(4π
k
)2 ∑
ℓmℓ′m′
ei(δℓ−δℓ′) sin δℓ sin δℓ′ Y
∗ℓm(k)Yℓ′m′(k)Yℓm(r)Y ∗
ℓ′m′(r), (49)
or, with k = kz,
dσ
dΩ(θ) =
1
k2
∑
ℓℓ′
(2ℓ+ 1)(2ℓ′ + 1)ei(δℓ−δℓ′) sin δℓ sin δℓ′ Pℓ(cos θ)Pℓ′(cos θ). (50)
These expressions have cross terms that do not simplify, but which show up experimentally as
oscillations in the angular dependence of the differential cross section.
Notes 35: Introduction to Scattering Theory 15
On the other hand, when we integrate over all angles to obtain the total cross section, the cross
terms integrate to zero. This is easiest to see in the form (49), due to the orthonormality of the
Yℓm’s. The result is
σ =
∫
dΩdσ
dΩ=
(4π
k
)2 ∑
ℓm
sin2 δℓ Y∗ℓm(k)Yℓm(k), (51)
or, with another application of the addition theorem (and noting that k · k = 1 and Pℓ(1) = 1),
σ =4π
k2
∑
ℓ
(2ℓ+ 1) sin2 δℓ. (52)
This is the partial wave expansion of the total cross section.
10. Hard Sphere Scattering
As an example of the partial wave expansion, let us consider scattering from a hard sphere.
The potential is
V (r) =∞, r < a,0, r > a,
(53)
where a is the radius of the sphere. In the region r > a the solution is a linear combination of the
j- and y-type spherical Bessel functions, which we write as
Rkℓ(r) = Ajℓ(kr) −Byℓ(kr), (54)
for some constants A and B. The radial wave function Rkℓ is labeled by both the angular momentum
ℓ and effective wave number k, where E = h2k2/2m. The wave function must vanish at r = a, so
we have
0 = Ajℓ(ka)−Byℓ(ka), (55)
which determines the ratio of B and A,
B
A=jℓ(ka)
yℓ(ka). (56)
To obtain the actual values of A and B we must specify a normalization convention for Rkℓ(r).
We will choose a normalization so that
A2 +B2 = 1, (57)
which specifies A and B to within a sign. We fix the sign by setting
A = −yℓ(ka)
D= cos δℓ,
B = −jℓ(ka)
D= sin δℓ,
(58)
where the denominator D is given by
D =√
jℓ(ka)2 + yℓ(ka)2. (59)
16 Notes 35: Introduction to Scattering Theory
Equation (58) also defines the angle δℓ modulo 2π, which will turn out to be the phase shift. With
these definitions, the solution in the region r > a is
Rkℓ(r) =1
Ddet
[
jℓ(ka) yℓ(ka)
jℓ(kr) yℓ(kr)
]
= cos δℓ jℓ(kr) − sin δℓ yℓ(kr). (60)
Now taking the asymptotic forms (16.29) and (16.30), valid when kr ≫ ℓ, we find
Rkℓ(r) ∼1
krsin
(
kr −ℓπ
2+ δℓ
)
, (61)
confirming that δℓ is the phase shift.
Knowledge of the δℓ implies knowledge of the scattering amplitude and differential cross section.
We will consider two limiting cases, in which the limiting forms of the spherical Bessel functions jℓ
and yℓ can be used.
11. The limit ka≪ 1
In the first case we assume ka ≪ 1. This is equivalent to λ ≫ a, where λ = 2π/k is the
wavelength of the incident waves, that is, the scatterer is much smaller than a wavelength. The
small argument forms of jℓ and yℓ, Eqs. (16.27) and (16.28), show that in this case yℓ(ka) is large
and negative and jℓ(ka) is small and positive. Then Eq. (58) shows that cos δℓ is near 1 and sin δℓ
is small and negative for all values of ℓ. Thus, to a good approximation,
sin δℓ ≈ δℓ ≈ −(ka)2ℓ+1
(2ℓ− 1)!!(2ℓ+ 1)!!. (62)
This shows that not only is the lowest phase shift δ0 small, the higher phase shifts δℓ for ℓ ≥ 1 are
even smaller. In this limit
δ0 = −ka, (63)
and the series (48) for the scattering amplitude is dominated by the single term ℓ = 0. Thus we
have
f(θ) =1
k(−ka) = −a, ka≪ 1, (64)
since eiδℓ ≈ 1 and P0(cos θ) = 1. We see that the differential cross section is independent of angle,
dσ
dΩ= a2, (65)
and the total cross section is
σ =
∫
dΩdσ
dΩ= 4πa2. (66)
The total cross section is four times as large as the geometrical cross-sectional area of the small
sphere. The classical cross section is just this cross-sectional area, πa2, but the regime λ≫ a is the
one in which we expect classical mechanics to be a poor approximation to quantum mechanics, so
we should not be surprised by the difference.
Notes 35: Introduction to Scattering Theory 17
12. s-wave Scattering
When the partial wave ℓ = 0 dominates the expansion of the scattering amplitude, we speak
of s-wave scattering. The scattered wave is isotropic and has an equal intensity in all directions,
including the forward direction. Since a ≪ λ the sphere is too small to create a shadow; instead,
the incident wave effectively wraps all the way around the scatterer. As we will see later, s-wave
scattering applies to any localized potential whose characteristic size a satisfies a ≪ λ, that is,
ka ≫ 1 (not just hard spheres). The fundamental reason is that the radial wave functions must
tunnel through the centrifugal potential to reach the scatterer, and the tunneling is deeper the higher
the ℓ value. Thus, the effect of the scatterer, as measured by the phase shift, is an exponentially
decreasing function of ℓ. Recall that the phase shift δℓ is a measure of the deviation from the
behavior of a free particle in the ℓ-th partial wave.
There are many physical examples in which s-wave scattering is important. For example, a Bose-
Einstein condensate is a dilute gas of atoms at a temperature at which the de Broglie wavelength
λ of the atoms due to their thermal motion is larger than the interparticle separation. Since the
gas is dilute, the interparticle separation in turn is much larger than the atomic size, call it a, so
we have λ ≫ a, or ka ≪ 1. Thus the interactions of the atoms with one another is described by
s-wave scattering, and scattering of atoms by atoms, in all its complexity, is described by a single
parameter, which is the phase shift δ0.
In s-wave scattering any two potentials that give the same phase shift δ0 will have the same
effect. Therefore in theoretical models the exact potential can be replaced by a simpler one, as long
as the phase shift is the same. Delta function potentials are popular for this purpose, which explains
their appearance in the models of Bose-Einstein condensates such as the Gross-Pitaevski equation.
When light scatters from particles that are much smaller than the wavelength, for example, when
optical radiation is scattered by atoms in a gas, is this s-wave scattering, and is the scattered wave
isotropic? (This is a tricky question.) The answer is no, because electromagnetic waves are vector
waves, not scalar waves as we have been discussing in these notes. In the case of electromagnetic
waves, the scattering when ka≪ 1 is dominated by dipole radiation, which has a nontrivial angular
dependence. Electromagnetic waves have no monopole radiation (that is, s-wave radiation), and the
next higher term in the multipole series is the first one to appear. It is the one that dominates at
small ka.
13. The Limit ka≫ 1
The second case we consider is the opposite extreme, ka ≫ 1, that is, λ ≪ a. In this case the
sphere is much larger than a wavelength. Now there are many ℓ values that contribute to the partial
wave sum. For small ℓ, for which ℓ ≪ ka, we can use the asymptotic forms for jℓ(ka) and yℓ(ka),
Eqs. (16.29) and (16.30). Substituting these into Eqs. (58) and (59) we find D = 1/ka and
δℓ = −ka+ℓπ
2. (67)
18 Notes 35: Introduction to Scattering Theory
This says that δℓ advances by π/2 when ℓ increases by 1, but this is an approximation that is only
valid when ℓ ≪ ka. As ℓ increases toward ka, the increment in δℓ gradually gets smaller. When ℓ
reaches and exceeds ka, the phase shifts δℓ rapidly approach zero, and the partial wave expansion
(48) effectively cuts off. The same is true for the expansion (52) of the total cross section.
a
Lmin
Lmax
Fig. 4. Particles of fixed energy are launched in various directions from a point on the surface of a sphere. Themaximum angular momentum is achieved when the direction is tangent to the sphere at the point of launch.
An intuitive explanation for the cutoff at ℓ ≈ ka is illustrated in Fig. 4. We imagine particles of
fixed energy E = h2k2/2m emitted in various directions from a point on the surface of a sphere of
radius a. These represent the waves scattered by the sphere. The angular momentum of the particles
depends on the direction of emission; its minimum value is L = 0 when the particles are launched
in the radial direction, and its maximum value L = pa = hka is when the particles are launched
tangentially to the sphere. The maximum value of L corresponds to a maximum angular momentum
quantum number, ℓ = L/h = ka. The same reasoning works for any short-range potential with an
effective range of a; the partial wave expansion cuts off near ℓ ≈ ka.
The phase shift δℓ is determined by Eqs. (58) to within a multiple of 2π. Thus each δℓ corre-
sponds to a point on a circle with coordinates x = cos δℓ, y = sin δℓ. As ℓ increases away from ℓ = 0,
this point advances by a phase angle around the circle, where for ℓ ≪ ka the increment is π/2, as
indicated by Eq. (67). As mentioned,this increment decreases as ℓ increases toward ka. Finally,
when ℓ ≈ ka, the phase shifts approach zero. This behavior is illustrated in Figs. 5 and 6.
While the points representing the δℓ are orbiting their circle as in Figs. 5 and 6, the average
value of sin2 δℓ is 1/2. Since the series cuts off near ℓ ≈ ka, the total cross section (52) can be
estimated by
σ ≈4π
k2
ka∑
ℓ=0
(2ℓ+ 1)1
2. (68)
Using the sumL∑
ℓ=0
(2ℓ+ 1) = (L+ 1)2, (69)
Notes 35: Introduction to Scattering Theory 19
15x
y
ka = 10
0
5
10
Fig. 5. Phase shifts in hard sphere scattering. The phaseshifts δℓ are represented as points on a circle with planepolar angle δℓ. The radius of the circle grows as ℓ increases,to keep the points from getting mixed up with each otheras they orbit the circle. Some points are labeled withtheir ℓ values. The phase shifts increase with decreasingincrement as ℓ increases, until ℓ ≈ ka, whereupon δℓ → 0.Plot is for ka = 10.
ka = 20
x
y
0 5
10
15
20
25
Fig. 6. Same as Fig. 5 except ka = 20.
we obtain the estimate
σ ≈4π
k2(ka)2
2= 2πa2, (70)
where we ignore the difference between ka and ka + 1. We see that the cross section is twice the
classical value in the case ka≫ 1.
These conclusions are confirmed by detailed calculations, which show that σ → 4πa2 as ka→ 0
and σ → 2πa2 as ka→ ∞. See Fig. 7.
If we think of the classical limit as one in which the de Broglie wavelength λ is much smaller
than the scale of the potential, then it is surprising that the limit σ → 2πa2 as ka → ∞ since this
is twice the classical cross section, πa2. To see how this “extra” cross-section is distributed in angle
we refer to Fig. 8, a plot of the differential cross section for various values of ka.
Figure 8 shows that for small values of ka, the differential cross section is nearly independent
of angle, confirming that we have s-wave scattering. The cross section dσ/dΩ is nearly a2 in this
range, confirming Eq. (65). As ka increases, dσ/dΩ decreases at large angles toward the classical
value, but at small angles a forward peak rises up, growing narrower and higher. At the edge of
forward peak there also arise oscillations. Figure 9 is a similar plot except that dσ/dΩ has been
multiplied by sin θ, which makes the integrated cross section proportional to area under the curve.
This makes it easier to see that the total cross section in the forward peak is about the same as the
classical cross section. In fact, detailed calculations (see Prob. 1) show that both the forward peak
and the classical cross section contribute πa2 to the total. The “extra” cross section comes entirely
from the forward peak.
This peak is due to diffraction. As the waves pass by the edge of the sphere they are bent
20 Notes 35: Introduction to Scattering Theory
0.01 0.1 1 10 100 1000
2.0
2.5
3.0
3.5
4.0
ka
σ/πa2
Fig. 7. Total cross section σ as a multiple of πa2 for hard sphere scattering as a function of ka. The total cross sectionhas the limits σ → 4πa2 as ka→ 0 and σ → 2πa2 as ka→ ∞.
0 90 1800
2
4
6
8
10
0.11
2
4
8
1
a2dσ
dΩ
θ
Fig. 8. Differential cross section in hard sphere scattering,as a multiple of a2. Curves are labeled by the value ofka, which increases from 0.1 to 8. Scattering angle θ ismeasured in degrees.
0 90 1800
1
2
3
4
0.112
4
8
16
1
a2dσ
dΩsin θ
θ
Fig. 9. Same as Fig. 8 except multiplied by sin θ. Areaunder curves in a given angle range is proportional to thetotal cross section in that range.
in an inward direction, that is, the direction necessary to close in the shadow behind the sphere.
Diffraction theory shows that if we move far enough downstream, the shadow is completely filled by
diffracted waves. Of course the differential cross section is defined in the limit r → ∞, so there is
Notes 35: Introduction to Scattering Theory 21
no shadow in the differential cross section. Diffraction causes the direction of the waves to change
from the incident direction, and so must be counted as part of the scattering.
14. The Optical Theorem
The optical theorem is an exact relationship between the total cross section σ and the scattering
amplitude f . In the context of potential scattering that we have considered in these notes, the optical
theorem is
σ =4π
kIm f(0), (71)
where f(0) refers to the scattering amplitude in the forward direction.
The optical theorem is trivial to prove in the case of scattering by central force potentials. We
place the z-axis along the direction of the beam so that the usual polar angle θ is the scattering
angle. Then Eq. (48) implies
4π
kIm f(0) =
4π
k2Im
∞∑
ℓ=0
(2ℓ+ 1)eiδℓ sin δℓ Pℓ(1) =4π
k2
∞∑
ℓ=0
(2ℓ+ 1) sin2 δℓ = σ, (72)
where we have used Eqs. (52) and (15.75). This proof is straightforward but it gives no insight into
what the theorem means.
The theorem (71) actually applies to scattering by any potential V (x) that dies off faster than
1/r as r → ∞, so that the asymptotic wave function is
ψ(x) ∼ eik·x +eikr
rf(θ, φ), (73)
as indicated by Eqs. (5) and (6). That is, we need not assume that the potential is rotationally
invariant. The two terms in Eq. (73) are the incident and scattered waves, respectively.
Since the wave function ψ(x) is a solution of the time-independent Schrodinger equation, the
current
J = Reψ∗(x)(
−ih∇
m
)
ψ(x) (74)
satisfies ∇ · J = 0, so by Stokes’ theorem the integral of J over any closed surface vanishes. See
Eqs. (5.55)–(5.57). We will integrate J over a large sphere of radius r centered at the origin of the
coordinates, which we assume is inside the region occupied by the potential as in Fig. 2. We will
take the limit r → ∞, so that the asymptotic form (73) of the wave function can be used.
Thus the asymptotic form of the current is
J ∼ Re[
e−ik·x +e−ikr
rf∗(θ, φ)
](
−ih∇
m
)[
eik·x +eikr
rf(θ, φ)
]
= vRe[
e−ikr cos θ +e−ikr
rf∗(θ, φ)
][
z eikr cos θ + reikr
rf(θ, φ)
]
,
(75)
where we have set k = kz and v = hk/m and retained only the leading term in the gradient of
the scattered wave, as in Eq. (11). The neglected terms will not contribute to the integral over the
sphere when we take the limit r → ∞.
22 Notes 35: Introduction to Scattering Theory
The current is quadratic in the wave function ψ(x), so in addition to the incident current
and scattered current there is a contribution coming from the cross terms, which represent the
interference between the incident wave and the scattered wave. We write
J = Jinc + Jscatt + Jx, (76)
where Jx is the interference current. By Stokes’ theorem we have
∫
sphere
J · da = 0, (77)
where da = r2r dΩ is the area element of the sphere.
Jinc
Fig. 10. The incident current passes straight through thesphere and produces no net flux across its surface.
Jscatt
Fig. 11. The scattered current produces a net flux ofparticles leaving the sphere, proportional to the total crosssection.
The incident current is
Jinc = vRe[
e−ikr cos θ][
z eikr cos θ]
= v z, (78)
It is a constant vector whose integral over sphere vanishes,
∫
sphere
Jinc · da = 0, (79)
as is obvious from Fig. 10.
The scattered current was already computed in Sec. 5. It is
Jscatt = vRe[e−ikr
rf∗(θ, φ)
][
reikr
rf(θ, φ)
]
= vr
r2|f(θ, φ)|2, (80)
whose integral over the sphere is
∫
sphere
J · da = v
∫
sphere
|f(θ, φ)|2 dΩ = σv, (81)
Notes 35: Introduction to Scattering Theory 23
where σ is the total cross section and where we have used Eq. (14). The scattered wave gives a net
outgoing flux, proportional to the cross section, which is no surprise. See Fig. 11.
Therefore the interference current must provide a net inward flux that cancels the scattered
flux. From Eq. (75) we have
Jx =v
rRe
[
z e−ikr(1−cos θ) f∗(θ, φ) + r eikr(1−cos θ) f(θ, φ)]
. (82)
But the real part of a complex number is the same as the real part of its complex conjugate, so we
can replace the first term in Eq. (82) by its complex conjugate to obtain
Jx =v
rRe
[
(z+ r) eikr(1−cos θ) f(θ, φ)]
. (83)
Now the interference flux is∫
sphere
Jx · da = vrRe
∫ 2π
0
dφ
∫ π
0
sin θ dθ (1 + cos θ) eikr(1−cos θ) f(θ, φ). (84)
We wish to evaluate this integral in the limit r → ∞, which is slightly delicate. The prefactor
of r would tend to make the integral diverge, but the factor eikr(1−cos θ) in the integrand oscillates
ever more rapidly in θ as r gets larger, thereby chopping up the integrand and tending to make the
integral vanish. To clarify the competition between these tendencies, we integrate by parts in θ,
obtaining
∫
sphere
Jx · da = vrRe
∫ 2π
0
dφ
∫ π
0
dθ1
ikr
[ d
dθeikr(1−cos θ)
]
(1 + cos θ)f(θ, φ)
=v
kRe
∫ 2π
0
dφ
−ieikr(1−cos θ) (1 + cos θ)f(θ, φ)∣
∣
∣
θ=π
θ=0
+ i
∫ π
0
dθ eikr(1−cos θ) d
dθ
[
(1 + cos θ)f(θ, φ)]
.
(85)
The integration by parts has eliminated the prefactor of r, so now the final integral oscillates itself
to death as r → ∞ and can be dropped. Evaluating the remaining term under the φ-integral at the
limits θ = 0, π, we see that it vanishes at θ = π and that at θ = 0 it is proportional to f(0, φ) which
is what we are writing simply as f(0) and which is independent of φ. Thus we find
∫
sphere
Jx · da =v
kRe
∫ 2π
0
dφ 2if(0) = −4πv
kIm f(0). (86)
The sum of this plus the scattered flux σv vanishes, which produces the optical theorem (71).
We see that the interference flux comes in from the forward direction, where it partially cancels
the incident flux going in the other direction.
The optical theorem has many generalizations. One version of it applies for the scattering
of classical electromagnetic waves (vector waves, in contrast to the scalar waves considered here).
Another applies to inelastic scattering in quantum mechanics, in which σ is the total cross section,
24 Notes 35: Introduction to Scattering Theory
including both elastic and inelastic scattering, while f(0) refers only to the forward amplitude for
elastic scattering. This is because only the wave for elastic scattering can interfere with the incident
wave. There are many other applications in fields ranging from quantum mechanics to black hole
physics.
Problems
1. The strange thing about scattering from a hard sphere in the limit ka≫ 1 is that the total cross
section is 2πr2, not πr2, the geometrical cross section. When the wave length is short, we expect
quantum mechanics to agree with classical mechanics, but it does not in this case.
(a) Work out the classical differential cross section dσ/dΩ for a hard sphere of radius a, and integrate
it to get the total cross section σ.
(b) In problem 9.1, you worked out the far field wave function ψ(x, y, z), for z ≫ ka2, when a plane
wave eikz traveling in the positive z-direction strikes a screen in the x-y plane with a circular hole
of radius a cut out. In that problem the hole was centered on the origin. The solution was worked
out for θ = ρ/z ≪ 1 (the paraxial approximation), where ρ =√
x2 + y2.
By subtracting this solution from the incident wave eikz , you get the far field wave function
when a plane wave eikz strikes the complementary screen, that is, just a disk of radius a at the
origin.
It turns out this wave field is the same as the wave field in hard sphere (of radius a) scattering
in the limit ka ≫ 1, if measured in the forward direction. That is because for forward scattering
from a hard scatterer, the physics is dominated by diffraction, so it is only the projection of the
scatterer onto the x-y plane that matters.
Write the scattered wave as (eikr/r)f(θ), express r as a function of z and θ for small θ, expand
out to lowest order in θ, and compare to the asymptotic wave field to get an expression for the
scattering amplitude for small angles θ. According to the optical theorem, the total cross section is
given in terms of the imaginary part of the forward scattering amplitude by
σ =4π
kIm f(0). (87)
Use this formula to compute σ.
(c) Show that dσ/dΩ in the forward direction has a narrow peak of width ∆θ ∼ 1/ka ≪ 1. Write
down an integral giving the contribution of this forward peak to the total cross section in terms
of the first root b of the Bessel function J1. You can approximate sin θ = θ in this integral, since
θ is small. It turns out that the value of this integral does not change much if the upper limit is
extended to infinity. Use the integral∫ ∞
0
dx
xJ1(x)
2 =1
2, (88)
Notes 35: Introduction to Scattering Theory 25
to find the contribution of the forward peak to the total cross section. (See Gradshteyn and Ryzhik,
integral number 6.538.2.)
2. This problem is borrowed from Sakurai. Consider a potential
V (r) =
V0 = const, r < a,
0, r > a,(89)
where V0 may be positive or negative. Using the method of partial waves, show that for |V0| ≪ E =
h2k2/2m and ka ≪ 1 the differential cross section is isotropic and that the total cross section is
given by
σtot =16π
9
m2V 20 a
6
h4. (90)
Suppose the energy is raised slightly. Show that the angular distribution can then be written as
dσ
dΩ= A+B cos θ. (91)
Obtain an approximate expression for B/A.